Refine
Has Fulltext
- yes (139)
Is part of the Bibliography
- yes (139)
Year of publication
Document Type
- Doctoral Thesis (139) (remove)
Language
- English (139) (remove)
Keywords
- Leistungsbewertung (29)
- Netzwerk (10)
- Robotik (10)
- Quality of Experience (9)
- Optimierung (7)
- Performance Evaluation (7)
- Autonomer Roboter (6)
- Cloud Computing (6)
- Komplexitätstheorie (6)
- Maschinelles Lernen (6)
Institute
- Institut für Informatik (139) (remove)
Schriftenreihe
Sonstige beteiligte Institutionen
- Deutsches Zentrum für Luft- und Raumfahrt (DLR), Institut für Raumfahrtsysteme (2)
- Zentrum für Telematik e.V. (2)
- Airbus Defence and Space GmbH (1)
- DLR (1)
- Deutsches Zentrum für Luft- und Raumfahrt e.V. (1)
- European Space Agency (1)
- Technische Hochschule Nürnberg Georg Simon Ohm (1)
- University of Applied Sciences and Arts Western Switzerland, Fribourg (1)
The ongoing and evolving usage of networks presents two critical challenges for current and future networks that require attention: (1) the task of effectively managing the vast and continually increasing data traffic and (2) the need to address the substantial number of end devices resulting from the rapid adoption of the Internet of Things. Besides these challenges, there is a mandatory need for energy consumption reduction, a more efficient resource usage, and streamlined processes without losing service quality. We comprehensively address these efforts, tackling the monitoring and quality assessment of streaming applications, a leading contributor to the total Internet traffic, as well as conducting an exhaustive analysis of the network performance within a Long Range Wide Area Network (LoRaWAN), one of the rapidly emerging LPWAN solutions.
The ongoing and evolving usage of networks presents two critical challenges for current and future networks that require attention: (1) the task of effectively managing the vast and continually increasing data traffic and (2) the need to address the substantial number of end devices resulting from the rapid adoption of the Internet of Things. Besides these challenges, there is a mandatory need for energy consumption reduction, a more efficient resource usage, and streamlined processes without losing service quality. We comprehensively address these efforts, tackling the monitoring and quality assessment of streaming applications, a leading contributor to the total Internet traffic, as well as conducting an exhaustive analysis of the network performance within a Long Range Wide Area Network (LoRaWAN), one of the rapidly emerging LPWAN solutions.
Wireless communication networks already comprise an integral part of both the private and industrial sectors and are successfully replacing existing wired networks. They enable the development of novel applications and offer greater flexibility and efficiency. Although some efforts are already underway in the aerospace sector to deploy wireless communication networks on board spacecraft, none of these projects have yet succeeded in replacing the hard-wired state-of-the-art architecture for intra-spacecraft communication. The advantages are evident as the reduction of the wiring harness saves time, mass, and costs, and makes the whole integration process more flexible. It also allows for easier scaling when interconnecting different systems.
This dissertation deals with the design and implementation of a wireless network architecture to enhance intra-spacecraft communications by breaking with the state-of-the-art standards that have existed in the space industry for decades. The potential and benefits of this novel wireless network architecture are evaluated, an innovative design using ultra-wideband technology is presented. It is combined with a Medium Access Control (MAC) layer tailored for low-latency and deterministic networks supporting even mission-critical applications. As demonstrated by the Wireless Compose experiment on the International Space Station (ISS), this technology is not limited to communications but also enables novel positioning applications.
To adress the technological challenges, extensive studies have been carried out on electromagnetic compatibility, space radiation, and data robustness. The architecture was evaluated from various perspectives and successfully demonstrated in space.
Overall, this research highlights how a wireless network can improve and potentially replace existing state-of-the-art communication systems on board spacecraft in future missions. And it will help to adapt and ultimately accelerate the implementation of wireless networks in space systems.
Deep Learning (DL) models are trained on a downstream task by feeding (potentially preprocessed) input data through a trainable Neural Network (NN) and updating its parameters to minimize the loss function between the predicted and the desired output. While this general framework has mainly remained unchanged over the years, the architectures of the trainable models have greatly evolved. Even though it is undoubtedly important to choose the right architecture, we argue that it is also beneficial to develop methods that address other components of the training process. We hypothesize that utilizing domain knowledge can be helpful to improve DL models in terms of performance and/or efficiency. Such model-agnostic methods can be applied to any existing or future architecture. Furthermore, the black box nature of DL models motivates the development of techniques to understand their inner workings. Considering the rapid advancement of DL architectures, it is again crucial to develop model-agnostic methods.
In this thesis, we explore six principles that incorporate domain knowledge to understand or improve models. They are applied either on the input or output side of the trainable model. Each principle is applied to at least two DL tasks, leading to task-specific implementations. To understand DL models, we propose to use Generated Input Data coming from a controllable generation process requiring knowledge about the data properties. This way, we can understand the model’s behavior by analyzing how it changes when one specific high-level input feature changes in the generated data. On the output side, Gradient-Based Attribution methods create a gradient at the end of the NN and then propagate it back to the input, indicating which low-level input features have a large influence on the model’s prediction. The resulting input features can be interpreted by humans using domain knowledge.
To improve the trainable model in terms of downstream performance, data and compute efficiency, or robustness to unwanted features, we explore principles that each address one of the training components besides the trainable model. Input Masking and Augmentation directly modifies the training input data, integrating knowledge about the data and its impact on the model’s output. We also explore the use of Feature Extraction using Pretrained Multimodal Models which can be seen as a beneficial preprocessing step to extract useful features. When no training data is available for the downstream task, using such features and domain knowledge expressed in other modalities can result in a Zero-Shot Learning (ZSL) setting, completely eliminating the trainable model. The Weak Label Generation principle produces new desired outputs using knowledge about the labels, giving either a good pretraining or even exclusive training dataset to solve the downstream task. Finally, improving and choosing the right Loss Function is another principle we explore in this thesis. Here, we enrich existing loss functions with knowledge about label interactions or utilize and combine multiple task-specific loss functions in a multitask setting.
We apply the principles to classification, regression, and representation tasks as well as to image and text modalities. We propose, apply, and evaluate existing and novel methods to understand and improve the model. Overall, this thesis introduces and evaluates methods that complement the development and choice of DL model architectures.
Graphs provide a key means to model relationships between entities.
They consist of vertices representing the entities,
and edges representing relationships between pairs of entities.
To make people conceive the structure of a graph,
it is almost inevitable to visualize the graph.
We call such a visualization a graph drawing.
Moreover, we have a straight-line graph drawing
if each vertex is represented as a point
(or a small geometric object, e.g., a rectangle)
and each edge is represented as a line segment between its two vertices.
A polyline is a very simple straight-line graph drawing,
where the vertices form a sequence according to which the vertices are connected by edges.
An example of a polyline in practice is a GPS trajectory.
The underlying road network, in turn, can be modeled as a graph.
This book addresses problems that arise
when working with straight-line graph drawings and polylines.
In particular, we study algorithms
for recognizing certain graphs representable with line segments,
for generating straight-line graph drawings,
and for abstracting polylines.
In the first part, we first examine,
how and in which time we can decide
whether a given graph is a stick graph,
that is, whether its vertices can be represented as
vertical and horizontal line segments on a diagonal line,
which intersect if and only if there is an edge between them.
We then consider the visual complexity of graphs.
Specifically, we investigate, for certain classes of graphs,
how many line segments are necessary for any straight-line graph drawing,
and whether three (or more) different slopes of the line segments
are sufficient to draw all edges.
Last, we study the question,
how to assign (ordered) colors to the vertices of a graph
with both directed and undirected edges
such that no neighboring vertices get the same color
and colors are ascending along directed edges.
Here, the special property of the considered graph is
that the vertices can be represented as intervals
that overlap if and only if there is an edge between them.
The latter problem is motivated by an application
in automated drawing of cable plans with vertical and horizontal line segments,
which we cover in the second part.
We describe an algorithm that
gets the abstract description of a cable plan as input,
and generates a drawing that takes into account
the special properties of these cable plans,
like plugs and groups of wires.
We then experimentally evaluate the quality of the resulting drawings.
In the third part, we study the problem of abstracting (or simplifying)
a single polyline and a bundle of polylines.
In this problem, the objective is to remove as many vertices as possible from the given polyline(s)
while keeping each resulting polyline sufficiently similar to its original course
(according to a given similarity measure).
In this thesis, we are interested in numerically preserving stationary solutions of balance laws. We start by developing finite volume well-balanced schemes for the system of Euler equations and the system of MHD equations with gravitational source term. Since fluid models and kinetic models are related, this leads us to investigate AP schemes for kinetic equations and their ability to preserve stationary solutions. Kinetic models typically have a stiff term, thus AP schemes are needed to capture good solutions of the model. For such kinetic models, equilibrium solutions are reached after large time. Thus we need a new technique to numerically preserve stationary solutions for AP schemes. We find a criterion for SP schemes for kinetic equations which states, that AP schemes under a particular discretization are also SP. In an attempt to mimic our result for kinetic equations in the context of fluid models, for the isentropic Euler equations we developed an AP scheme in the limit of the Mach number going to zero. Our AP scheme is proven to have a SP property under the condition that the pressure is a function of the density and the latter is obtained as a solution of an elliptic equation. The properties of the schemes we developed and its criteria are validated numerically by various test cases from the literature.
Environmental issues have emerged especially since humans burned fossil fuels, which led to air pollution and climate change that harm the environment. These issues’ substantial consequences evoked strong efforts towards assessing the state of our environment.
Various environmental machine learning (ML) tasks aid these efforts. These tasks concern environmental data but are common ML tasks otherwise, i.e., datasets are split (training, validatition, test), hyperparameters are optimized on validation data, and test set metrics measure a model’s generalizability. This work focuses on the following environmental ML tasks: Regarding air pollution, land use regression (LUR) estimates air pollutant concentrations at locations where no measurements are available based on measured locations and each location’s land use (e.g., industry, streets). For LUR, this work uses data from London (modeled) and Zurich (measured). Concerning climate change, a common ML task is model output statistics (MOS), where a climate model’s output for a study area is altered to better fit Earth observations and provide more accurate climate data. This work uses the regional climate model (RCM) REMO and Earth observations from the E-OBS dataset for MOS. Another task regarding climate is grain size distribution interpolation where soil properties at locations without measurements are estimated based on the few measured locations. This can provide climate models with soil information, that is important for hydrology. For this task, data from Lower Franconia is used.
Such environmental ML tasks commonly have a number of properties: (i) geospatiality, i.e., their data refers to locations relative to the Earth’s surface. (ii) The environmental variables to estimate or predict are usually continuous. (iii) Data can be imbalanced due to relatively rare extreme events (e.g., extreme precipitation). (iv) Multiple related potential target variables can be available per location, since measurement devices often contain different sensors. (v) Labels are spatially often only sparsely available since conducting measurements at all locations of interest is usually infeasible. These properties present challenges but also opportunities when designing ML methods for such tasks.
In the past, environmental ML tasks have been tackled with conventional ML methods, such as linear regression or random forests (RFs). However, the field of ML has made tremendous leaps beyond these classic models through deep learning (DL). In DL, models use multiple layers of neurons, producing increasingly higher-level feature representations with growing layer depth. DL has made previously infeasible ML tasks feasible, improved the performance for many tasks in comparison to existing ML models significantly, and eliminated the need for manual feature engineering in some domains due to its ability to learn features from raw data. To harness these advantages for environmental domains it is promising to develop novel DL methods for environmental ML tasks.
This thesis presents methods for dealing with special challenges and exploiting opportunities inherent to environmental ML tasks in conjunction with DL. To this end, the proposed methods explore the following techniques: (i) Convolutions as in convolutional neural networks (CNNs) to exploit reoccurring spatial patterns in geospatial data. (ii) Posing the problems as regression tasks to estimate the continuous variables. (iii) Density-based weighting to improve estimation performance for rare and extreme events. (iv) Multi-task learning to make use of multiple related target variables. (v) Semi–supervised learning to cope with label sparsity. Using these techniques, this thesis considers four research questions: (i) Can air pollution be estimated without manual feature engineering? This is answered positively by the introduction of the CNN-based LUR model MapLUR as well as the off-the-shelf LUR solution OpenLUR. (ii) Can colocated pollution data improve spatial air pollution models? Multi-task learning for LUR is developed for this, showing potential for improvements with colocated data. (iii) Can DL models improve the quality of climate model outputs? The proposed DL climate MOS architecture ConvMOS demonstrates this. Additionally, semi-supervised training of multilayer perceptrons (MLPs) for grain size distribution interpolation is presented, which can provide improved input data. (iv) Can DL models be taught to better estimate climate extremes? To this end, density-based weighting for imbalanced regression (DenseLoss) is proposed and applied to the DL architecture ConvMOS, improving climate extremes estimation. These methods show how especially DL techniques can be developed for environmental ML tasks with their special characteristics in mind. This allows for better models than previously possible with conventional ML, leading to more accurate assessment and better understanding of the state of our environment.
Serverless computing is an emerging cloud computing paradigm that offers a highlevel
application programming model with utilization-based billing. It enables the
deployment of cloud applications without managing the underlying resources or
worrying about other operational aspects. Function-as-a-Service (FaaS) platforms
implement serverless computing by allowing developers to execute code on-demand
in response to events with continuous scaling while having to pay only for the
time used with sub-second metering. Cloud providers have further introduced
many fully managed services for databases, messaging buses, and storage that also
implement a serverless computing model. Applications composed of these fully
managed services and FaaS functions are quickly gaining popularity in both industry
and in academia.
However, due to this rapid adoption, much information surrounding serverless
computing is inconsistent and often outdated as the serverless paradigm evolves.
This makes the performance engineering of serverless applications and platforms
challenging, as there are many open questions, such as: What types of applications
is serverless computing well suited for, and what are its limitations? How should
serverless applications be designed, configured, and implemented? Which design
decisions impact the performance properties of serverless platforms and how can
they be optimized? These and many other open questions can be traced back to an
inconsistent understanding of serverless applications and platforms, which could
present a major roadblock in the adoption of serverless computing.
In this thesis, we address the lack of performance knowledge surrounding serverless
applications and platforms from multiple angles: we conduct empirical studies
to further the understanding of serverless applications and platforms, we introduce
automated optimization methods that simplify the operation of serverless applications,
and we enable the analysis of design tradeoffs of serverless platforms by
extending white-box performance modeling.
The landscape of today’s programming languages is manifold. With the diversity of applications, the difficulty of adequately addressing and specifying the used programs increases. This often leads to newly designed and implemented domain-specific languages. They enable domain experts to express knowledge in their preferred format, resulting in more readable and concise programs. Due to its flexible and declarative syntax without reserved keywords, the logic programming language Prolog is particularly suitable for defining and embedding domain-specific languages.
This thesis addresses the questions and challenges that arise when integrating domain-specific languages into Prolog. We compare the two approaches to define them either externally or internally, and provide assisting tools for each. The grammar of a formal language is usually defined in the extended Backus–Naur form. In this work, we handle this formalism as a domain-specific language in Prolog, and define term expansions that allow to translate it into equivalent definite clause grammars. We present the package library(dcg4pt) for SWI-Prolog, which enriches them by an additional argument to automatically process the term’s corresponding parse tree. To simplify the work with definite clause grammars, we visualise their application by a web-based tracer.
The external integration of domain-specific languages requires the programmer to keep the grammar, parser, and interpreter in sync. In many cases, domain-specific languages can instead be directly embedded into Prolog by providing appropriate operator definitions. In addition, we propose syntactic extensions for Prolog to expand its expressiveness, for instance to state logic formulas with their connectives verbatim. This allows to use all tools that were originally written for Prolog, for instance code linters and editors with syntax highlighting. We present the package library(plammar), a standard-compliant parser for Prolog source code, written in Prolog. It is able to automatically infer from example sentences the required operator definitions with their classes and precedences as well as the required Prolog language extensions. As a result, we can automatically answer the question: Is it possible to model these example sentences as valid Prolog clauses, and how?
We discuss and apply the two approaches to internal and external integrations for several domain-specific languages, namely the extended Backus–Naur form, GraphQL, XPath, and a controlled natural language to represent expert rules in if-then form. The created toolchain with library(dcg4pt) and library(plammar) yields new application opportunities for static Prolog source code analysis, which we also present.
There is great interest in affordable, precise and reliable metrology underwater:
Archaeologists want to document artifacts in situ with high detail.
In marine research, biologists require the tools to monitor coral growth and geologists need recordings to model sediment transport.
Furthermore, for offshore construction projects, maintenance and inspection millimeter-accurate measurements of defects and offshore structures are essential.
While the process of digitizing individual objects and complete sites on land is well understood and standard methods, such as Structure from Motion or terrestrial laser scanning, are regularly applied, precise underwater surveying with high resolution is still a complex and difficult task.
Applying optical scanning techniques in water is challenging due to reduced visibility caused by turbidity and light absorption.
However, optical underwater scanners provide significant advantages in terms of achievable resolution and accuracy compared to acoustic systems.
This thesis proposes an underwater laser scanning system and the algorithms for creating dense and accurate 3D scans in water.
It is based on laser triangulation and the main optical components are an underwater camera and a cross-line laser projector.
The prototype is configured with a motorized yaw axis for capturing scans from a tripod.
Alternatively, it is mounted to a moving platform for mobile mapping.
The main focus lies on the refractive calibration of the underwater camera and laser projector, the image processing and 3D reconstruction.
For highest accuracy, the refraction at the individual media interfaces must be taken into account.
This is addressed by an optimization-based calibration framework using a physical-geometric camera model derived from an analytical formulation of a ray-tracing projection model.
In addition to scanning underwater structures, this work presents the 3D acquisition of semi-submerged structures and the correction of refraction effects.
As in-situ calibration in water is complex and time-consuming, the challenge of transferring an in-air scanner calibration to water without re-calibration is investigated, as well as self-calibration techniques for structured light.
The system was successfully deployed in various configurations for both static scanning and mobile mapping.
An evaluation of the calibration and 3D reconstruction using reference objects and a comparison of free-form surfaces in clear water demonstrate the high accuracy potential in the range of one millimeter to less than one centimeter, depending on the measurement distance.
Mobile underwater mapping and motion compensation based on visual-inertial odometry is demonstrated using a new optical underwater scanner based on fringe projection.
Continuous registration of individual scans allows the acquisition of 3D models from an underwater vehicle.
RGB images captured in parallel are used to create 3D point clouds of underwater scenes in full color.
3D maps are useful to the operator during the remote control of underwater vehicles and provide the building blocks to enable offshore inspection and surveying tasks.
The advancing automation of the measurement technology will allow non-experts to use it, significantly reduce acquisition time and increase accuracy, making underwater metrology more cost-effective.
Deep learning enables enormous progress in many computer vision-related tasks. Artificial Intel- ligence (AI) steadily yields new state-of-the-art results in the field of detection and classification. Thereby AI performance equals or exceeds human performance. Those achievements impacted many domains, including medical applications.
One particular field of medical applications is gastroenterology. In gastroenterology, machine learning algorithms are used to assist examiners during interventions. One of the most critical concerns for gastroenterologists is the development of Colorectal Cancer (CRC), which is one of the leading causes of cancer-related deaths worldwide. Detecting polyps in screening colonoscopies is the essential procedure to prevent CRC. Thereby, the gastroenterologist uses an endoscope to screen the whole colon to find polyps during a colonoscopy. Polyps are mucosal growths that can vary in severity.
This thesis supports gastroenterologists in their examinations with automated detection and clas- sification systems for polyps. The main contribution is a real-time polyp detection system. This system is ready to be installed in any gastroenterology practice worldwide using open-source soft- ware. The system achieves state-of-the-art detection results and is currently evaluated in a clinical trial in four different centers in Germany.
The thesis presents two additional key contributions: One is a polyp detection system with ex- tended vision tested in an animal trial. Polyps often hide behind folds or in uninvestigated areas. Therefore, the polyp detection system with extended vision uses an endoscope assisted by two additional cameras to see behind those folds. If a polyp is detected, the endoscopist receives a vi- sual signal. While the detection system handles the additional two camera inputs, the endoscopist focuses on the main camera as usual.
The second one are two polyp classification models, one for the classification based on shape (Paris) and the other on surface and texture (NBI International Colorectal Endoscopic (NICE) classification). Both classifications help the endoscopist with the treatment of and the decisions about the detected polyp.
The key algorithms of the thesis achieve state-of-the-art performance. Outstandingly, the polyp detection system tested on a highly demanding video data set shows an F1 score of 90.25 % while working in real-time. The results exceed all real-time systems in the literature. Furthermore, the first preliminary results of the clinical trial of the polyp detection system suggest a high Adenoma Detection Rate (ADR). In the preliminary study, all polyps were detected by the polyp detection system, and the system achieved a high usability score of 96.3 (max 100). The Paris classification model achieved an F1 score of 89.35 % which is state-of-the-art. The NICE classification model achieved an F1 score of 81.13 %.
Furthermore, a large data set for polyp detection and classification was created during this thesis. Therefore a fast and robust annotation system called Fast Colonoscopy Annotation Tool (FastCAT) was developed. The system simplifies the annotation process for gastroenterologists. Thereby the
i
gastroenterologists only annotate key parts of the endoscopic video. Afterward, those video parts are pre-labeled by a polyp detection AI to speed up the process. After the AI has pre-labeled the frames, non-experts correct and finish the annotation. This annotation process is fast and ensures high quality. FastCAT reduces the overall workload of the gastroenterologist on average by a factor of 20 compared to an open-source state-of-art annotation tool.
Since the first CubeSat launch in 2003, the hardware and software complexity of the nanosatellites was continuosly increasing.
To keep up with the continuously increasing mission complexity and to retain the primary advantages of a CubeSat mission, a new approach for the overall space and ground software architecture and protocol configuration is elaborated in this work.
The aim of this thesis is to propose a uniform software and protocol architecture as a basis for software development, test, simulation and operation of multiple pico-/nanosatellites based on ultra-low power components.
In contrast to single-CubeSat missions, current and upcoming nanosatellite formation missions require faster and more straightforward development, pre-flight testing and calibration procedures as well as simultaneous operation of multiple satellites.
A dynamic and decentral Compass mission network was established in multiple active CubeSat missions, consisting of uniformly accessible nodes.
Compass middleware was elaborated to unify the communication and functional interfaces between all involved mission-related software and hardware components.
All systems can access each other via dynamic routes to perform service-based M2M communication.
With the proposed model-based communication approach, all states, abilities and functionalities of a system are accessed in a uniform way.
The Tiny scripting language was designed to allow dynamic code execution on ultra-low power components as a basis for constraint-based in-orbit scheduler and experiment execution.
The implemented Compass Operations front-end enables far-reaching monitoring and control capabilities of all ground and space systems.
Its integrated constraint-based operations task scheduler allows the recording of complex satellite operations, which are conducted automatically during the overpasses.
The outcome of this thesis became an enabling technology for UWE-3, UWE-4 and NetSat CubeSat missions.
A graph is an abstract network that represents a set of objects, called vertices, and relations between these objects, called edges. Graphs can model various networks. For example, a social network where the vertices correspond to users of the network and the edges represent relations between the users. To better see the structure of a graph it is helpful to visualize it. A standard visualization is a node-link diagram in the Euclidean plane. In such a representation the vertices are drawn as points in the plane and edges are drawn as Jordan curves between every two vertices connected by an edge. Edge crossings decrease the readability of a drawing, therefore, Crossing Optimization is a fundamental problem in Computer Science. This book explores the research frontiers and introduces novel approaches in Crossing Optimization.
An enduring engineering problem is the creation of unreliable software leading to unreliable systems. One reason for this is source code is written in a complicated manner making it too hard for humans to review and understand. Complicated code leads to other issues beyond dependability, such as expanded development efforts and ongoing difficulties with maintenance, ultimately costing developers and users more money.
There are many ideas regarding where blame lies in the reation of buggy and unreliable systems. One prevalent idea is the selected life cycle model is to blame. The oft-maligned “waterfall” life cycle model is a particularly popular recipient of blame. In response, many organizations changed their life cycle model in hopes of addressing these issues. Agile life cycle models have become very popular, and they promote communication between team members and end users. In theory, this communication leads to fewer misunderstandings and should lead to less complicated and more reliable code.
Changing the life cycle model can indeed address communications ssues, which can resolve many problems with understanding requirements.
However, most life cycle models do not specifically address coding practices or software architecture. Since lifecycle models do not address the structure of the code, they are often ineffective at addressing problems related to code complicacy.
This dissertation answers several research questions concerning software complicacy, beginning with an investigation of traditional metrics and static analysis to evaluate their usefulness as measurement tools. This dissertation also establishes a new concept in applied linguistics by creating a measurement of software complicacy based on linguistic economy. Linguistic economy describes the efficiencies of speech, and this thesis shows the applicability of linguistic economy to software. Embedded in each topic is a discussion
of the ramifications of overly complicated software, including the relationship of complicacy to software faults. Image recognition using machine learning is also investigated as a potential method of identifying problematic source code.
The central part of the work focuses on analyzing the source code of hundreds of different projects from different areas. A static analysis was performed on the source code of each project, and traditional software metrics were calculated. Programs were also analyzed using techniques developed by linguists to measure expression and statement complicacy and identifier complicacy. Professional software engineers were also directly surveyed to understand mainstream perspectives.
This work shows it is possible to use traditional metrics as indicators of potential project bugginess. This work also discovered it is possible to use image recognition to identify problematic pieces of source code. Finally, this work discovered it is possible to use linguistic methods to determine which statements and expressions are least desirable and more complicated for programmers.
This work’s principle conclusion is that there are multiple ways to discover traits indicating a project or a piece of source code has characteristics of being buggy. Traditional metrics and static analysis can be used to gain some understanding of software complicacy and bugginess potential. Linguistic economy demonstrates a new tool for measuring software complicacy, and machine learning can predict where bugs may lie in source code. The significant implication of this work is developers can recognize when a project is becoming buggy and take practical steps to avoid creating buggy projects.
Detecting anomalies in transaction data is an important task with a high potential to avoid financial loss due to irregularities deliberately or inadvertently carried out, such as credit card fraud, occupational fraud in companies or ordering and accounting errors. With ongoing digitization of our world, data-driven approaches, including machine learning, can draw benefit from data with less manual effort and feature engineering. A large variety of machine learning-based anomaly detection methods approach this by learning a precise model of normality from which anomalies can be distinguished. Modeling normality in transactional data, however, requires to capture distributions and dependencies within the data precisely with special attention to numerical dependencies such as quantities, prices or amounts.
To implicitly model numerical dependencies, Neural Arithmetic Logic Units have been proposed as neural architecture. In practice, however, these have stability and precision issues.
Therefore, we first develop an improved neural network architecture, iNALU, which is designed to better model numerical dependencies as found in transaction data. We compare this architecture to the previous approach and show in several experiments of varying complexity that our novel architecture provides better precision and stability.
We integrate this architecture into two generative neural network models adapted for transaction data and investigate how well normal behavior is modeled. We show that both architectures can successfully model normal transaction data, with our neural architecture improving generative performance for one model.
Since categorical and numerical variables are common in transaction data, but many machine learning methods only process numerical representations, we explore different representation learning techniques to transform categorical transaction data into dense numerical vectors. We extend this approach by proposing an outlier-aware discretization, thus incorporating numerical attributes into the computation of categorical embeddings, and investigate latent spaces, as well as quantitative performance for anomaly detection.
Next, we evaluate different scenarios for anomaly detection on transaction data. We extend our iNALU architecture to a neural layer that can model both numerical and non-numerical dependencies and evaluate it in a supervised and one-class setting. We investigate the stability and generalizability of our approach and show that it outperforms a variety of models in the balanced supervised setting and performs comparably in the one-class setting. Finally, we evaluate three approaches to using a generative model as an anomaly detector and compare the anomaly detection performance.
Latency is an inherent problem of computing systems. Each computation takes time until the result is available. Virtual reality systems use elaborated computer resources to create virtual experiences. The latency of those systems is often ignored or assumed as small enough to provide a good experience.
This cumulative thesis is comprised of published peer reviewed research papers exploring the behaviour and effects of latency. Contrary to the common description of time invariant latency, latency is shown to fluctuate. Few other researchers have looked into this time variant behaviour. This thesis explores time variant latency with a focus on randomly occurring latency spikes. Latency spikes are observed both for small algorithms and as end to end latency in complete virtual reality systems. Most latency measurements gather close to the mean latency with potentially multiple smaller clusters of larger latency values and rare extreme outliers. The latency behaviour differs for different implementations of an algorithm. Operating system schedulers and programming language environments such as garbage collectors contribute to the overall latency behaviour. The thesis demonstrates these influences on the example of different implementations of message passing.
The plethora of latency sources result in an unpredictable latency behaviour. Measuring and reporting it in scientific experiments is important. This thesis describes established approaches to measuring latency and proposes an enhanced setup to gather detailed information. The thesis proposes to dissect the measured data with a stacked z-outlier-test to separate the clusters of latency measurements for better reporting.
Latency in virtual reality applications can degrade the experience in multiple ways. The thesis focuses on cybersickness as a major detrimental effect. An approach to simulate time variant latency is proposed to make latency available as an independent variable in experiments to understand latency's effects. An experiment with modified latency shows that latency spikes can contribute to cybersickness. A review of related research shows that different time invariant latency behaviour also contributes to cybersickness.
With the miniaturization of satellites a fundamental change took place in the space industry. Instead of single big monolithic satellites nowadays more and more systems are envisaged consisting of a number of small satellites to form cooperating systems in space. The lower costs for development and launch as well as the spatial distribution of these systems enable the implementation of new scientific missions and commercial services.
With this paradigm shift new challenges constantly emerge for satellite developers, particularly in the area of wireless communication systems and network protocols.
Satellites in low Earth orbits and ground stations form dynamic space-terrestrial networks. The characteristics of these networks differ fundamentally from those of other networks.
The resulting challenges with regard to communication system design, system analysis, packet forwarding, routing and medium access control as well as challenges concerning the reliability and efficiency of wireless communication links are addressed in this thesis.
The physical modeling of space-terrestrial networks is addressed by analyzing existing satellite systems and communication devices, by evaluating measurements and by implementing a simulator for space-terrestrial networks.
The resulting system and channel models were used as a basis for the prediction of the dynamic network topologies, link properties and channel interference. These predictions allowed for the implementation of efficient routing and medium access control schemes for space-terrestrial networks. Further, the implementation and utilization of software-defined ground stations is addressed, and a data upload scheme for the operation of small satellite formations is presented.
Human-computer interfaces have the potential to support mental health practitioners in alleviating mental distress.
Adaption of this technology in practice is, however, slow.
We provide means to extend the design space of human-computer interfaces for mitigating mental distress.
To this end, we suggest three complementary approaches: using presentation technology, using virtual environments, and using communication technology to facilitate social interaction.
We provide new evidence that elementary aspects of presentation technology affect the emotional processing of virtual stimuli, that perception of our environment affects the way we assess our environment, and that communication technologies affect social bonding between users.
By showing how interfaces modify emotional reactions and facilitate social interaction, we provide converging evidence that human-computer interfaces can help alleviate mental distress.
These findings may advance the goal of adapting technological means to the requirements of mental health practitioners.
The importance of proactive and timely prediction of critical events is steadily increasing, whether in the manufacturing industry or in private life. In the past, machines in the manufacturing industry were often maintained based on a regular schedule or threshold violations, which is no longer competitive as it causes unnecessary costs and downtime. In contrast, the predictions of critical events in everyday life are often much more concealed and hardly noticeable to the private individual, unless the critical event occurs. For instance, our electricity provider has to ensure that we, as end users, are always supplied with sufficient electricity, or our favorite streaming service has to guarantee that we can watch our favorite series without interruptions. For this purpose, they have to constantly analyze what the current situation is, how it will develop in the near future, and how they have to react in order to cope with future conditions without causing power outages or video stalling.
In order to analyze the performance of a system, monitoring mechanisms are often integrated to observe characteristics that describe the workload and the state of the system and its environment. Reactive systems typically employ thresholds, utility functions, or models to determine the current state of the system. However, such reactive systems cannot proactively estimate future events, but only as they occur. In the case of critical events, reactive determination of the current system state is futile, whereas a proactive system could have predicted this event in advance and enabled timely countermeasures. To achieve proactivity, the system requires estimates of future system states. Given the gap between design time and runtime, it is typically not possible to use expert knowledge to a priori model all situations a system might encounter at runtime. Therefore, prediction methods must be integrated into the system. Depending on the available monitoring data and the complexity of the prediction task, either time series forecasting in combination with thresholding or more sophisticated machine and deep learning models have to be trained.
Although numerous forecasting methods have been proposed in the literature, these methods have their advantages and disadvantages depending on the characteristics of the time series under consideration. Therefore, expert knowledge is required to decide which forecasting method to choose. However, since the time series observed at runtime cannot be known at design time, such expert knowledge cannot be implemented in the system. In addition to selecting an appropriate forecasting method, several time series preprocessing steps are required to achieve satisfactory forecasting accuracy. In the literature, this preprocessing is often done manually, which is not practical for autonomous computing systems, such as Self-Aware Computing Systems. Several approaches have also been presented in the literature for predicting critical events based on multivariate monitoring data using machine and deep learning. However, these approaches are typically highly domain-specific, such as financial failures, bearing failures, or product failures. Therefore, they require in-depth expert knowledge. For this reason, these approaches cannot be fully automated and are not transferable to other use cases. Thus, the literature lacks generalizable end-to-end workflows for modeling, detecting, and predicting failures that require only little expert knowledge.
To overcome these shortcomings, this thesis presents a system model for meta-self-aware prediction of critical events based on the LRA-M loop of Self-Aware Computing Systems. Building upon this system model, this thesis provides six further contributions to critical event prediction. While the first two contributions address critical event prediction based on univariate data via time series forecasting, the three subsequent contributions address critical event prediction for multivariate monitoring data using machine and deep learning algorithms. Finally, the last contribution addresses the update procedure of the system model. Specifically, the seven main contributions of this thesis can be summarized as follows:
First, we present a system model for meta self-aware prediction of critical events. To handle both univariate and multivariate monitoring data, it offers univariate time series forecasting for use cases where a single observed variable is representative of the state of the system, and machine learning algorithms combined with various preprocessing techniques for use cases where a large number of variables are observed to characterize the system’s state. However, the two different modeling alternatives are not disjoint, as univariate time series forecasts can also be included to estimate future monitoring data as additional input to the machine learning models. Finally, a feedback loop is incorporated to monitor the achieved prediction quality and trigger model updates.
We propose a novel hybrid time series forecasting method for univariate, seasonal time series, called Telescope. To this end, Telescope automatically preprocesses the time series, performs a kind of divide-and-conquer technique to split the time series into multiple components, and derives additional categorical information. It then forecasts the components and categorical information separately using a specific state-of-the-art method for each component. Finally, Telescope recombines the individual predictions. As Telescope performs both preprocessing and forecasting automatically, it represents a complete end-to-end approach to univariate seasonal time series forecasting. Experimental results show that Telescope achieves enhanced forecast accuracy, more reliable forecasts, and a substantial speedup. Furthermore, we apply Telescope to the scenario of predicting critical events for virtual machine auto-scaling. Here, results show that Telescope considerably reduces the average response time and significantly reduces the number of service level objective violations.
For the automatic selection of a suitable forecasting method, we introduce two frameworks for recommending forecasting methods. The first framework extracts various time series characteristics to learn the relationship between them and forecast accuracy. In contrast, the other framework divides the historical observations into internal training and validation parts to estimate the most appropriate forecasting method. Moreover, this framework also includes time series preprocessing steps. Comparisons between the proposed forecasting method recommendation frameworks and the individual state-of-the-art forecasting methods and the state-of-the-art forecasting method recommendation approach show that the proposed frameworks considerably improve the forecast accuracy.
With regard to multivariate monitoring data, we first present an end-to-end workflow to detect critical events in technical systems in the form of anomalous machine states. The end-to-end design includes raw data processing, phase segmentation, data resampling, feature extraction, and machine tool anomaly detection. In addition, the workflow does not rely on profound domain knowledge or specific monitoring variables, but merely assumes standard machine monitoring data. We evaluate the end-to-end workflow using data from a real CNC machine. The results indicate that conventional frequency analysis does not detect the critical machine conditions well, while our workflow detects the critical events very well with an F1-score of almost 91%.
To predict critical events rather than merely detecting them, we compare different modeling alternatives for critical event prediction in the use case of time-to-failure prediction of hard disk drives. Given that failure records are typically significantly less frequent than instances representing the normal state, we employ different oversampling strategies. Next, we compare the prediction quality of binary class modeling with downscaled multi-class modeling. Furthermore, we integrate univariate time series forecasting into the feature generation process to estimate future monitoring data. Finally, we model the time-to-failure using not only classification models but also regression models. The results suggest that multi-class modeling provides the overall best prediction quality with respect to practical requirements. In addition, we prove that forecasting the features of the prediction model significantly improves the critical event prediction quality.
We propose an end-to-end workflow for predicting critical events of industrial machines. Again, this approach does not rely on expert knowledge except for the definition of monitoring data, and therefore represents a generalizable workflow for predicting critical events of industrial machines. The workflow includes feature extraction, feature handling, target class mapping, and model learning with integrated hyperparameter tuning via a grid-search technique. Drawing on the result of the previous contribution, the workflow models the time-to-failure prediction in terms of multiple classes, where we compare different labeling strategies for multi-class classification. The evaluation using real-world production data of an industrial press demonstrates that the workflow is capable of predicting six different time-to-failure windows with a macro F1-score of 90%. When scaling the time-to-failure classes down to a binary prediction of critical events, the F1-score increases to above 98%.
Finally, we present four update triggers to assess when critical event prediction models should be re-trained during on-line application. Such re-training is required, for instance, due to concept drift. The update triggers introduced in this thesis take into account the elapsed time since the last update, the prediction quality achieved on the current test data, and the prediction quality achieved on the preceding test data. We compare the different update strategies with each other and with the static baseline model. The results demonstrate the necessity of model updates during on-line application and suggest that the update triggers that consider both the prediction quality of the current and preceding test data achieve the best trade-off between prediction quality and number of updates required.
We are convinced that the contributions of this thesis constitute significant impulses for the academic research community as well as for practitioners. First of all, to the best of our knowledge, we are the first to propose a fully automated, end-to-end, hybrid, component-based forecasting method for seasonal time series that also includes time series preprocessing. Due to the combination of reliably high forecast accuracy and reliably low time-to-result, it offers many new opportunities in applications requiring accurate forecasts within a fixed time period in order to take timely countermeasures. In addition, the promising results of the forecasting method recommendation systems provide new opportunities to enhance forecasting performance for all types of time series, not just seasonal ones. Furthermore, we are the first to expose the deficiencies of the prior state-of-the-art forecasting method recommendation system.
Concerning the contributions to critical event prediction based on multivariate monitoring data, we have already collaborated closely with industrial partners, which supports the practical relevance of the contributions of this thesis. The automated end-to-end design of the proposed workflows that do not demand profound domain or expert knowledge represents a milestone in bridging the gap between academic theory and industrial application. Finally, the workflow for predicting critical events in industrial machines is currently being operationalized in a real production system, underscoring the practical impact of this thesis.
In today's world, circumstances, processes, and requirements for systems in general-in this thesis a special focus is given to the context of Cyber-Physical Systems (CPS)-are becoming increasingly complex and dynamic.
In order to operate properly in such dynamic environments, systems must adapt to dynamic changes, which has led to the research area of Self-Adaptive Systems (SAS).
These systems can deal with changes in their environment and the system itself.
In our daily lives, we come into contact with many different self-adaptive systems that are designed to support and improve our way of life.
In this work we focus on the two domains Intelligent Transportation Systems (ITS) and logistics as both domains provide complex and adaptable use cases to prototypical apply the contributions of this thesis.
However, the contributions are not limited to these areas and can be generalized also to other domains such as the general area of CPS and Internet of Things including smart grids or even intelligent computer networks.
In ITS, real-time traffic control is an example adaptive system that monitors the environment, analyzes observations, and plans and executes adaptation actions.
Another example is platooning, which is the ability of vehicles to drive with close inter-vehicle distances.
This technology enables an increase in road throughput and safety, which directly addresses the increased infrastructure needs due to increased traffic on the roads.
In logistics, the Vehicle Routing Problem (VRP) deals with the planning of road freight transport tours.
To cope with the ever-increasing transport volume due to the rise of just-in-time production and online shopping, efficient and correct route planning for transports is important.
Further, warehouses play a central role in any company's supply chain and contribute to the logistical success.
The processes of storage assignment and order picking are the two main tasks in mezzanine warehouses highly affected by a dynamic environment.
Usually, optimization algorithms are applied to find solutions in reasonable computation time.
SASes can help address these dynamics by allowing systems to deal with changing demands and constraints.
For the application of SASes in the two areas ITS and logistics, the definition of adaptation planning strategies is the key success factor.
A wide range of adaptation planning strategies for different domains can be found in the literature, and the operator must select the most promising strategy for the problem at hand.
However, the No-Free-Lunch theorem states that the performance of one strategy is not necessarily transferable to other problems.
Accordingly, the algorithm selection problem, first defined in 1976, aims to find the best performing algorithm for the current problem.
Since then, this problem has been explored more and more, and the machine learning community, for example, considers it a learning problem.
The ideas surrounding the algorithm selection problem have been applied in various use cases, but little research has been done to generalize the approaches.
Moreover, especially in the field of SASes, the selection of the most appropriate strategy depends on the current situation of the system.
Techniques for identifying the situation of a system can be found in the literature, such as the use of rules or clustering techniques.
This knowledge can then be used to improve the algorithm selection, or in the scope of this thesis, to improve the selection of adaptation planning strategies.
In addition, knowledge about the current situation and the performance of strategies in similar previously observed situations provides another opportunity for improvements.
This ongoing learning and reasoning about the system and its environment is found in the research area Self-Aware Computing (SeAC).
In this thesis, we explore common characteristics of adaptation planning strategies in the domain of ITS and logistics presenting a self-aware optimization framework for adaptation planning strategies.
We consider platooning coordination strategies from ITS and optimization techniques from logistics as adaptation planning strategies that can be exchanged during operation to better reflect the current situation.
Further, we propose to integrate fairness and uncertainty handling mechanisms directly into the adaptation planning strategies.
We then examine the complex structure of the logistics use cases VRP and mezzanine warehouses and identify their systems-of-systems structure.
We propose a two-stage approach for vertical or nested systems and propose to consider the impact of intertwining horizontal or coexisting systems.
More specifically, we summarize the six main contributions of this thesis as follows:
First, we analyze specific characteristics of adaptation planning strategies with a particular focus on ITS and logistics.
We use platooning and route planning in highly dynamic environments as representatives of ITS and we use the rich Vehicle Routing Problem (rVRP) and mezzanine warehouses as representatives of the logistics domain.
Using these case studies, we derive the need for situation-aware optimization of adaptation planning strategies and argue that fairness is an important consideration when applying these strategies in ITS.
In logistics, we discuss that these complex systems can be considered as systems-of-systems and this structure affects each subsystem.
Hence, we argue that the consideration of these characteristics is a crucial factor for the success of the system.
Second, we design a self-aware optimization framework for adaptation planning strategies.
The optimization framework is abstracted into a third layer above the application and its adaptation planning system, which allows the concept to be applied to a diverse set of use cases.
Further, the Domain Data Model (DDM) used to configure the framework enables the operator to easily apply it by defining the available adaptation planning strategies, parameters to be optimized, and performance measures.
The framework consists of four components: (i) Coordination, (ii) Situation Detection, (iii) Strategy Selection, and (iv) Parameter Optimization.
While the coordination component receives observations and triggers the other components, the situation detection applies rules or clustering techniques to identify the current situation.
The strategy selection uses this knowledge to select the most promising strategy for the current situation, and the parameter optimization applies optimization algorithms to tune the parameters of the strategy.
Moreover, we apply the concepts of the SeAC domain and integrate learning and reasoning processes to enable ongoing advancement of the framework.
We evaluate our framework using the platooning use case and consider platooning coordination strategies as the adaptation planning strategies to be selected and optimized.
Our evaluation shows that the framework is able to select the most appropriate adaptation strategy and learn the situational behavior of the system.
Third, we argue that fairness aspects, previously identified as an important characteristic of adaptation planning strategies, are best addressed directly as part of the strategies.
Hence, focusing on platooning as an example use case, we propose a set of fairness mechanisms to balance positive and negative effects of platooning among all participants in a platoon.
We design six vehicle sequence rotation mechanisms that continuously change the leader position among all participants, as this is the position with the least positive effects.
We analyze these strategies on roads of different sizes and with different traffic volumes, and show that these mechanisms should also be chosen wisely.
Fourth, we address the uncertainty characteristic of adaptation planning strategies and propose a methodology to account for uncertainty and also address it directly as part of the adaptation planning strategies.
We address the use case of fueling planning along a route associated with highly dynamic fuel prices and develop six utility functions that account for different aspects of route planning.
Further, we incorporate uncertainty measures for dynamic fuel prices by adding penalties for longer travel times or greater distance to the next gas station.
Through this approach, we are able to reduce the uncertainty at planning time and obtain a more robust route planning.
Fifth, we analyze optimization of nested systems-of-systems for the use case rVRP.
Before proposing an approach to deal with the complex structure of the problem, we analyze important constraints and objectives that need to be considered when formulating a real-world rVRP.
Then, we propose a two-stage workflow to optimize both systems individually, flexibly, and interchangeably.
We apply Genetic Algorithms and Ant Colony Optimization (ACO) to both nested systems and compare the performance of our workflow with state-of-the-art optimization algorithms for this use case.
In our evaluation, we show that the proposed two-stage workflow is able to handle the complex structure of the problem and consider all real-world constraints and objectives.
Finally, we study coexisting systems-of-systems by optimizing typical processes in mezzanine warehouses.
We first define which ergonomic and economic constraints and objectives must be considered when addressing a real-world problem.
Then, we analyze the interrelatedness of the storage assignment and order picking problems; we identify opportunities to design optimization approaches that optimize all objectives and aim for a good overall system performance, taking into account the interdependence of both systems.
We use the NSGA-II for storage assignment and Ant Colony Optimization (ACO) for order picking and adapt them to the specific requirements of horizontal systems-of-systems.
In our evaluation, we compare our approaches to state-of-the-art approaches in mezzanine warehouses and show that our proposed approaches increase the system performance.
Our proposed approaches provide important contributions to both academic research and practical applications.
To the best of our knowledge, we are the first to design a self-aware optimization framework for adaptation planning strategies that integrates situation-awareness, algorithm selection, parameter tuning, as well as learning and reasoning.
Our evaluation of platooning coordination shows promising results for the application of the framework.
Moreover, our proposed strategies to compensate for negative effects of platooning represent an important milestone, which could lead to higher acceptance of this technology in society and support its future adoption in the real world.
The proposed methodology and utility functions that address uncertainty are an important step to improving the capabilities of SAS in an increasingly turbulent environment.
Similarly, our contributions to systems-of-systems optimization are major contributions to the state of logistics and systems-of-systems research.
Finally, we select real-world use cases for the application of our approaches and cooperate with industrial partners, which highlights the practical relevance of our contributions.
The reduction of manual effort and required expert knowledge in our self-aware optimization framework is a milestone in bridging the gap between academia and practice.
One of our partners integrated the two-stage approach to tackling the rVRP into its software system, improving both time to solution and solution quality.
In conclusion, the contributions of this thesis have spawned several research projects such as a long-term industrial project on optimizing tours and routes in parcel delivery funded by Bayerisches Verbundforschungsprogramm (BayVFP) – Digitalisierung and further collaborations, opening up many promising avenues for future research.